Newspapers in Times of Low Adversiting Revenues∗
Charles Angelucci†and Julia Cagé‡
Columbia University and Sciences Po Paris
November 4, 2015
Abstract
Newspaper subscribers are typically charged a lower per issue price than occasional
buyers. This difference in prices has recently increased. This paper posits that this tendency may be interpreted as a response to the continuing drop in advertising revenues.
We investigate how the reliance on advertising revenues interacts with the incentives
newspapers have to adopt subscriber-based readerships. More precisely, we investigate
theoretically and empirically the determinants of second-degree price discrimination in
two-sided markets. We build a model in which a newspaper must attract both readers and advertisers. Readers are uncertain as to their future benefit from reading, and
heterogeneous in their taste for reading. Advertisers are heterogeneous in their outside
option, taste for subscribers, and taste for occasional buyers. To estimate empirically
the effect of the advertisers’ side of the industry on price discrimination on the readers’
side, we use a “quasi-natural experiment”. We exploit the introduction of advertisement on French Television in 1968, which we treat as a negative shock on advertising
revenues of daily national newspapers (treated group), but not on daily local newspapers (control group). We build a new dataset on French local newspapers between
1960 and 1974 and perform a Differences-in-Differences analysis. We find robust evidence of increased price discrimination as a result of a drop in advertising revenues.
Keywords: price discrimination; two-sided markets; newspaper markets; advertising
JEL: C33, L11, M13
∗
This paper was previously circulated under the title “Price Discrimination in a Two-Sided Market: Theory
and Evidence from the Newspaper Industry”. We gratefully acknowledge the many helpful comments and
suggestions from Ariel Pakes, Valeria Rueda, Andrei Veiga, Glen Weyl and Alex White. We are also grateful
to seminar participants at Harvard University, INSEAD, the London School of Economics, Namur University,
the Paris School of Economics, Sciences Po Paris, SHUFE in Shanghai, Stockholm University and Warwick
University. Maikol Cerda and Charlotte Coutand provided outstanding research assistance. We gratefully
acknowledge financial support from the NET Institute (www.NETinst.org) and the Paris School of Economics.
All errors remain our own.
†
Columbia University, ca2630 [at] columbia [dot] edu.
‡
Sciences Po Paris, julia [dot] cage [at] sciencespo [dot] fr.
1
1
Introduction
The newspaper industry is a canonical example of a two-sided market: newspapers serve two
distinct groups of consumers – readers and advertisers, where each group cares about the
presence and characteristics of the other. The resulting network effects lead to subtle pricing
policies that have received much attention recently (see for instance Rochet and Tirole, 2003;
Weyl, 2010). One feature of newspapers’ pricing policies is the observed price discrimination
between subscribers and occasional buyers; subscribers are typically charged a lower per
issue price than occasional buyers, and these price differences appear not to be explained
by cost differences entirely. Furthermore, this difference in prices has recently increased and
newspapers tend now to favor a more subscriber-based readership. This paper posits that this
tendency may be interpreted as a response to the industry’s state of distress, itself in part
attributed to the continuing drop in adverting revenues. In the United States for example,
we indeed observe a decline in newspaper advertising revenues (as a share of GDP) since the
second half of the 1950’s, decline that has been sharper since the beginning of the 2000’s
(Figure 11 ). We also observe a decrease in newspapers’ reliance on advertising revenues (the
0
.2
% GDP
.4
.6
.8
share of advertising in newspaper total revenues).
1950
1960
1970
1980
1990
2000
2010
Notes: This Figure represents the evolution of newspaper advertising revenues as a share of GDP in the United States
between 1950 and 2013. Data on newspaper revenues is from the Newspaper Association of America (NAA). GDP data
is from the World Development Indicators (WDI).
Figure 1: Newspaper advertising revenues as a share of GDP in the United States, 1950-2103
In this paper we investigate how the reliance on advertising revenues interacts with the
incentives newspapers have to adopt subscriber-based readerships. To this end, we first extend recent models of multi-sided industries to incorporate the scope for second-degree price
1
Figure B.1 in the online Appendix represents the evolution of newspaper advertising revenues in the United
States over the same period in billion dollars.
2
discrimination between subscribers and occasional buyers. We then carry out an empirical
analysis using a “quasi-natural experiment” and a new dataset on the French local newspaper
industry that we build from archives data.
We build a general model of a two-sided market in which a monopolist newspaper repeatedly interacts with a continuum of readers and a continuum of advertisers. Newspapers can
be purchased by readers either by subscription or at the newsstand on a day-by-day basis.
Independently of the presence of advertisers, the scope for price discrimination stems from
(i) the readers’ uncertainty regarding their exact willingness to pay in future periods and (ii)
the readers’ heterogeneity in their average willingness to pay. Readers with a high average
willingness to pay subscribe at a low per unit price, while others buy the newspaper at a high
price whenever their willingness to pay is high.2
Advertisers are heterogenous in (i) their taste for subscribers, (ii) their taste for occasional readers, and (iii) their outside option (i.e., their payoff when placing ads on alternative
platforms). The challenge is to disentangle how the presence of advertisers affects the prices
charged to readers. We characterize the optimal pricing formulas of the newspaper, as well
as the readers and advertisers’ demands. These formulas are intuitive and in the spirit of
Weyl (2010). When choosing its prices, aside from taking into account the various marginal
costs and demand elasticities, the newspaper must cater to (i) the average taste of marginal
readers – those indifferent between subscribing or buying occasionally on the one hand, and
those indifferent between buying occasionally or never on the other – and (ii) the average
taste of marginal advertisers for both subscribers and non subscribers, as well as their outside
options.
On the empirical side, the main empirical challenge is to isolate the “advertising revenue”
effect on price discrimination. To this end, we follow an empirical strategy in the spirit of an
event study. We exploit the introduction of advertisement on French Television in October
1968 by treating it as an exogenous negative shock on the advertising side of newspapers.
Television is state-owned in France from 1945 to 1981. The introduction of advertisement
on television was decided by law, despite strong resistances by the newspaper industry. This
introduction leads to an exogenous shock that shifts exclusively the incentives to price discriminate stemming from advertising revenues. Indeed, reader heterogeneity and the various
marginal costs of producing and delivering newspapers were not affected. To the best of our
knowledge, we are the first to use this “quasi-natural” experiment.3
Our identifying assumption is that the negative shock on advertising revenues has affected
mostly national daily newspapers, and to a lower extent local daily newspapers. Indeed, while
2
This rationale for price discrimination was first introduced by Glazer and Hassin (1982), but in a model
without advertisers.
3
Filistrucchi et al. (2012) considers the “reverse” experiment: they analyze the effects of the advertising
ban on French public television in 2009. They find that it did not favour private TV channels at the expense
of public ones. They do not investigate how it affects newspapers nor price discrimination, however.
3
national newspaper advertisement consists mostly of commercial advertisements that are relatively close substitutes to those broadcasted on television (national brands, etc), a large share
of advertisements in local newspapers is instead local in nature (local commercial advertisements and classified advertisements). We document a substitution effect of advertisements
from national (but not local) newspapers to television by studying the actual content of the
advertisements broadcasted on television and of the advertisements published in newspapers
before and after the introduction of advertisement on TV.
National newspapers, for which national ads are likely to provide a significant fraction of
revenue, are more likely to respond to the negative shock on national advertising than local
newspaper, for which national ad-revenue is likely not their major source of revenue. We
thus use national newspapers as our “treated group”, and local newspapers as our “control
group”. Using novel annual data on local and national newspapers between 1960 and 1974,
we compare the pre-1968-to-post-1968 change in price discrimination by national daily newspapers to the change in price discrimination by local daily newspapers over the same period
(Difference-in-Difference estimation). We show that national newspapers, due to their greater
reliance on national-ad revenue, experience a larger drop in advertising rates and advertising
revenues than local newspapers after the introduction of advertisement on television. This
drop propagates to the reader side of the newspaper market. We find that the decrease in
advertising revenues leads to an increase in the extent of price discrimination, i.e., newspapers adopt a more subscriber-based readership as a consequence of the drop in advertising
revenues. This increase in price discrimination is driven by a decrease in the price charged
to subscribers. As a consequence, we find that the number of subscribers and the share of
subscribers in the total circulation increase after the shock. Our results are robust to a range
of alternative specifications and controls.
Our proposed interpretation of these empirical findings is that, while newspapers have
a preference for subscribers – having more subscribers allow them to increase their total
circulation, and to build consumer loyalty –, advertisers tend to prefer unit buyers. As a consequence, when advertising revenues play an important role, newspapers tend to distort their
pricing strategies in favor of unit buyers. On the contrary, when advertising revenues fall (e.g.
after the 1968 French reform), they revert to their own profit-maximizing pricing strategies
(independent of the advertising side of the market), i.e. they increase price discrimination in
favor of subscribers.
Various explanations may help rationalize advertisers’ preference for unit buyers. First, the
newsstand helps subscribers increase their reach: whereas a subscription goes to one person
or one household, a newsstand copy of a newspaper reaches more people. Next, consumers
buying at the newsstand have a real interest in the newspaper – the newsstand buyer is
probably likely to spend more time reading the newspaper than the comparable subscriber.
4
While these points have been raised by a number of consultants in the industry, to the extent
of our knowledge, we are the first to show empirically that it is indeed the case. Moreover,
the theoretical predictions of our model are in line with these preferences.
Our findings may have important implications for the 21st century newspaper industry. In
particular they shed light on the observed current tendency for newspapers to favor subscriberbased readerships through low subscription prices. According to our results, this tendency may
be rationalised by the continuous decline in advertising revenues as a result of the competition
stemming from Internet.4 Additionally, to the extent that suscribers and unit buyers are
different social and political groups (an issue that we do not directly adress in this paper),
rising price discimination could contribute to a more polarized access to information and to
diverging turnout rates by socio-economic groups.
Literature review This paper builds on the seminal model by Glazer and Hassin (1982)
who first study price discrimination by newspapers based on consumers uncertainty.5 We
introduce the advertising side in the profit function of newspapers and discuss how the reliance
on advertising revenues interacts with price discrimination on the reader side. Moreover, we
estimate this interaction empirically. This relates our paper to the growing literature that
examines empirically the determinants of price discrimination. A number of papers investigate
the role of competition. Seminal contributions include Borenstein (1991) on retail gasoline
markets and Borenstein and Rose (1994) on airline tickets. More recent articles include Busse
and Rysman (2005) who investigate pricing in Yellow pages advertising, Gerardi and Shapiro
(2009) who reexamine air ticket price discrimination, Dai et al. (2012) who study the nonmonotonicity of the effect of competition on price discrimination using data from the U.S.
airline, and Seim and Viard (2011) who study nonlinear pricing in cellular telecommunication
markets. To the exception of Gil and Riera-Crichton (2011) who empirically tests the relation
between price discrimination and competition in the two-sided market setting of the Spanish
local television industry, all theses articles study one-sided markets.6 Our paper aims on the
contrary at understanding the consequences of network effects on price discrimination.
Carroni (2015) provides a model of behaviour-based price discrimination – firms offer
different prices to consumers according to their past purchase behaviour – in the context
4
While Athey et al. (2013) explore theoretically the extent to which greater consumer switching facilitated
by online consumption of news explains the recent collapse in advertising revenue; Gentzkow (2014) investigate
how Internet has reduced the return news outlets can earn by selling the attention of their consumers to
advertisers.
5
On the economics of subscription sales, see also Gabszewicz and Sonnac (1997); Morton and Oster (2003);
Resende and Ferioli (2014).
6
Marcus Asplund Rickard Eriksson (2008), using evidence from Swedish newspapers, show that more competitive markets have a higher incidence on third-degree price discrimination. But they abstract from the
advertising side of the newspaper market.
5
of two-sided markets.7 There also exists a very recent vein of research that examines the
role of consumers’ bounded rationality on price discrimination via subscription (see Grubb,
2012, for an insightful review). Prominent contributions to this literature are DellaVigna and
Malmendier (2004) for contracts in health sport centers and Grubb (2009) for cellular phone
service plans. Although we recognize that bounded rationality may play a role in consumers’
decision as to subscribe or not to a newspaper, the scope for price discrimination in our
model instead stems from informational considerations. In addition, our aggregated data do
not allow us to investigate this issue.
Our paper also relates to the literature on two-sided markets. Rochet and Tirole (2003)
provide a widely applicable model of two-sided markets and discuss markets for advertising,
credit cards, software and web portal usage. Weyl (2010) and White and Weyl (2010) further
extend two-sided market models.8 Naturally, much work on two-sided markets has focused on
the media industry. Argentesi and Filistrucchi (2007) develop an analysis to estimate market
power in the Italian newspaper industry, but do not consider price discrimination on the reader
side, just like Kaiser and Wright (2006) who estimate a model of competition in the German
market for magazine readership and avertising. Seamans and Zhu (2014) look at the impact
of the entry of Craig’s list on local newspapers’ pricing policies and find that this negative
shock on the advertisement side of newspapers had led to an increase in subscription prices, a
result consistent with previous theoretical studies that show that an increase in competition
on one side of a two-sided market can lead to an increase in price on the other side (Godes
et al., 2009; Hagiu, 2009).
We contribute to this recent line of research by introducing price discrimination on one side
of the market. Seamans and Zhu (2014) find an increase in subscription price as a result of the
negative shock on the advertisement side; on the contrary, we obtain that the subscription
price decreases. But while they only focus on subscribers, we also take into account unit
buyers, which helps rationalize our finding. Our paper makes an important contribution by
investigating advertisers’ preferences for unit buyers compared to subscribers. Advertisers’
preference for unit buyers explains why newspapers react to the negative shock by decreasing
the subscription price. Price discrimination matters on the reader side. To the best of
our knowledge, Liu and Serfes (2013) is the only paper investigating price discrimination in
two-sided markets. However, their modeling approach does not fit well with the newspaper
industry as they consider perfect price discrimination on both sides in a Hotelling framework.
Our paper is the first theoretical and empirical analysis of second-degree price discrimination
7
For a dynamic model of duopolistic competition under behaviour-based price discrimination, but in onesided setting, see Caillaud and Nijs (2014).
8
Anderson and Coate (2000) study broadcast markets in which retailers pay for advertising to reach consumers, and where consumers dislike advertising. Rysman (2004) provides an empirical analysis of the market
for yellow pages. Jin and Rysman (2013) study US sports card conventions pricing though the lens of the
two-sided market theory.
6
in a two-sided market.
Finally, this research project is a contribution to the empirical literature on media using
historical data to understand the evolution of the newspaper industry and its impact on
society. In a paper investigating the effect of newspapers entry and exit, Gentzkow et al.
(2011) find that newspaper entry has a robust positive effect on political participation in the
United States. Cagé (2014) finds in the French context that competition has a negative effect.
In their recent work on competition and ideological diversity Gentzkow et al. (2014) estimate
a model of newspaper demand, entry, and political affiliation choice, in which newspapers
compete for both readers and advertisers. In this paper, we use historical data and a quasinatural experiment to estimate empirically the effect of the advertisers’ side of the industry
on price discrimination on the readers’ side.
The remainder of the paper is organized as follows. Section 2 develops a model of seconddegree price discrimination by a platform. Section 3 introduces the new dataset we built
for this study and provides descriptive statistics. In Section 4, we discuss the historical
context of the introduction of advertisement on French Television in 1968, and document a
substitution effect from national newspapers advertisements to advertisements broadcasted
on television. In Section 5 we estimate the effect of the advertising side of newspapers on
price discrimination on the reader side using a Differences-in-Differences analysis based on
the introduction of advertisement on French Television. Section 6 concludes.
2
Two-sided markets and second-degree price discrimination
We build a simple model of a two-sided market in which the platform (i.e., the newspaper)
engages in second-degree price discrimination on one side of the market.
2.1
Set-up
Consider the interaction between a monopoly newspaper, a continuum of readers of mass one
(side R of the industry), and a continuum of advertisers of mass one (side A). The newspaper
produces n distinct issues during the period of interest. Let t denote a given date/issue, where
t = 1, ..., n. Readers can either subscribe to the newspaper to receive all n issues, or buy some
issues on occasions. We refer to S as the subset of readers who choose to subscribe, and O as
the subset of readers who choose to be occasional readers.
Newspaper The profit-maximizing monopoly newspaper chooses price pi to charge group
i = O, S, A. We denote by ci the per-issue newspaper’s (constant) marginal cost of serving
group i. We do not model the actual production of news, and thus implicitly assume that the
newspaper produces content that is of interest to at least some readers.
7
Readers
The payoff to reader i from reading the newspaper at date t is given by:
R
Ui,t
= θi − γi N A + t − pi ,
(1)
where θi denotes reader i’s taste for reading, γi N A denotes reader i’s taste/distaste for the
quantity N A of ads in the newspaper, and where t captures a shock common to all readers
at date t (e.g., the holding of elections). We assume the pair (θi , γi ) is drawn according to
joint pdf f R (θ, γ), where each parameter has support going from minus to plus infinity.
Furthermore, t is drawn according to U [, ] and is i.i.d. across dates/issues. At each
date t = 1, ..., n, reader i, if she has not subscribed, observes the realization of t before
deciding whether to purchase the newspaper. Not subscribing therefore allows readers to
make informed purchasing decisions. By contrast, readers who choose to subscribe make
their decision prior to observing all n future realizations of . We assume readers have an
outside option equal to 0.
Let UiS denote reader i’s expected payoff when subscribing. For a given per-unit subscription price pS , it is equal to:9
UiS
=
n
X
Pr θi − γi N A + t ≥ 0 E θi − γi N A + t | θi − γi N A + t ≥ 0 − npS ,
t=1
2
n max θi + − γi N A , 0
− npS .
=
2
( − )
A subscriber does not necessarily read all n issues of the newspaper. Instead, Pr θi − γi N A + t ≥ 0
measures the probability that a given subscriber i reads the newspaper on a given t. The
price pS plays no role in this decision because it is already sunk by the time the decision to
read an issue comes. Naturally, though, a necessary condition for reader i to subscribe in the
first place is that (2.1) is nonnegative.
The expected payoff of a reader who chooses to be an occasional reader is equal to:
UiO =
n
X
Pr θi − γi N A + t − pO ≥ 0 E θi − γi N A + t − pO | θi − γi N A + t − pO ≥ 0 ,
t=1
(2)
NA
=
−
n max θi + − γi
2
( − )
2
pO , 0
.
Not surprisingly, the decision to purchase an issue of the newspaper depends on the price
pO . We refer to occasional readers who make a purchase with positive probability as “active”
occasional readers.
Anticipating on the computation of the demand functions, note that for some readers to
9
We assume readers do not discount future payoffs to simplify expressions.
8
subscribe, it must be the case that the per-unit subscription price
pO . If pO >
pS
n
pS
n
is lower than the price
, all readers choose not to subscribe (i) to enjoy the lower per-unit price pO
and (ii) to retain the ability not to purchase the newspaper on some dates. In the following,
we let ΘS denote the subset of readers who choose to subscribe, and ΘO the subset of readers
who choose to be “active” occasional readers. Also, ΘS (γ) denotes the subset of readers who
choose to subscribe within the set of readers with parameter γ, and ΘO (γ) denotes the subset
of active occasional readers within the set of readers with parameter γ (recall that readers are
heterogeneous on two dimensions).
If, at the optimum, the pair of prices pS , pO is such that some readers choose to become
subscribers and others to be active occasional readers—i.e., the case of empirical interest,
then, within the set of readers with parameter γ, the readers who choose to subscribe are
such that their taste for reading θ is higher than or equal to:
θ̃1 (θ) = γN A +
pS
pO
+
− .
2
( − ) pO
(3)
The threshold θ̃1 (θ) is increasing in γ, increasing in pS , and decreasing in pO . Readers who
decide to subscribe are those with high taste for reading: those who would often purchase
the newspaper at price pO but are given the opportunity to buy it at a lower price through
the subscription plan.10 By contrast, occasional readers are those with an average taste for
reading: they are willing to purchase the newspaper only when the shock is sufficiently high.
Advertisers
We assume advertisers choose between either (i) placing an ad in the newspa-
per for the n issues at price npA , or (ii) never placing an ad. The expected payoff to advertiser
j of placing an ad in the newspaper is equal to:
UjA
=
n X
h i
h
i
O
bSj E N̂tS + bO
E
N̂
− npA ,
j
t
(4)
t=1
where
∞
h i Z
E N̂tS =
−∞
Z
∞
=
−∞
Z
f R (γ, θ) Pr t ≥ γi N A − θi dθi dγi
− γ i N A − θi
f (γ, θ)
−
ΘS (γi )
Z
(5)
ΘS (γi )
R
dθi dγi ,
10
The highest number of subscribers the newspaper can attract is achieved by setting pS = 0. However, even
when pS = 0 not all readers choose to subscribe. Specifically, all readers whose payoffs are such that they can
derive positive utility from reading for at least some values of choose to subscribe, and the rest is indifferent
between subscribing (but never actually reading an issue) and never buying an issue. In other words, if pS = 0,
all “active” readers—those such that θ − γN A + ≥ 0 —choose to subscribe.
9
and where
∞
h
i Z
E N̂tO =
Z
−∞
Z
∞
f R (γ, θ) Pr t ≥ γi N A + pO − θi dθi dγi
− γ i N A + p O − θi
f (γ, θ)
−
ΘO (γi )
Z
=
−∞
(6)
ΘO (γi )
R
dθi dγi .
h i
E N̂ S denotes the expected number of subscribers who actually read the newspaper, ∀t.
h
i
Similarly, E N̂ O denotes the expected number of active occasional readers, ∀t. Advertisers
only care about readers who actually read the newspaper. Specifically, the parameter bSj
captures advertiser j’s taste for active subscribers and bO
j captures advertiser j’s taste for
active occasional readers. In addition, we assume advertiser j has outside option uA
j . The
A
A bS , bO , uA , where each parameter is
3-tuple (bSj , bO
j , uj ) is drawn according to joint pdf f
drawn from support going from minus infinity to plus infinity. We assume away any price
discrimination by the newspaper on the advertisers’ side of the industry. Advertiser j places
an ad in the newspaper for n issues at price npA if and only if:
n X
i
h
h i
O
− npA ≥ uA
bSj E N̂ S + bO
j .
j E N̂
(7)
t=1
2.2
The newspaper’s problem
We first compute the three relevant demand functions; that is, the demand for subscriptions,
the expected demand per issue at the newsstand, and the demand for advertising slots.
Demand functions
The number of subscribers is equal to:
N
S
S
O
A
p ,p ,N
Z
∞
Z
=
−∞
f R (γ, θ) dθi dγi .
(8)
ΘS (γi )
The number of subscribers is determined by the subscription price pS , the newsstand price
pO , and the number of ads N A . Similarly, the expected number of occasional readers per-issue
is equal to:
N
O
O
S
p ,p ,N
A
Z
∞
=
−∞
− γ i N A + p O − θi
f (γ, θ)
−
ΘO (γi )
Z
R
dθi dγi .
(9)
The expected number of occasional readers is determined by the subscription price pS , the
newsstand price pO , and the number of ads N A . On the other side of the industry, the number
10
of ads is equal to:
N
A
h
E N̂
S
i
h
, E N̂
O
i
A
,p
Z
∞
Z
O
A
Z n(bSj E [N̂ S ]+bO
j E [N̂ ])−np
∞
=
−∞
−∞
S O
f bS , bO , uA duA
j dbj dbj .
−∞
(10)
The number of ads is determined by the ad price pA , the expected number of active subscribers
h i
h
i
E N̂ S , and the expected number of active occasional readers E N̂ O .
Under regularity conditions, the system (8)-(9)-(10) has a unique solution characterizing
demands as a function of pO , pS , pA :
DS = dS pO , pS , pA ,
DO = dO pO , pS , pA ,
DA = dA pO , pS , pA ,
where DO is the expected demand stemming from occasional readers, ∀t.
The newspaper’s profit function is given by:
Π = ΠO + ΠS + ΠA
(11)
= n pO − c0 dO pO , pS , pA + n pS − cS dS pO , pS , pA + n pA − cA dA pO , pS , pA .
Note that, unlike advertisers, the newspaper cares about the total number of subscribers,
as opposed to the active number of subscribers. In the following, let ipj be the price pj
elasticity of demand di , where i, j = O, S, A. The next proposition states the newspaper’s
optimal pricing rule in case, at the optimum, the sets ΘS and ΘO are nonempty—that is, the
case of empirical interest. The formulas are obtained by rearranging the first-order conditions
of the newspaper’s problem.
Proposition 1 If it is optimal for the newspaper to sell its content both to subscribers and
occasional readers, the newspaper prices are given by the following augmented mark-up formulas:
pO = cO
S
S
p =c
pA = cA
O
pO
1 + O
pO
SpS
1 + SpS
A
pA
1 + A
pA
− pS − cS
O
O
− p −c
SpO
1 + O
pO
− pO − cO
O
pS
1 + SpS
O
pA
1 + A
pA
A
dS
dA
pO
A
A
−
p
−
c
dO
dO
1 + O
pO
A
dO
dA
pS
A
A
−
p
−
c
dS
1 + SpS dS
SpA dS
dO
S
S
−
p
−
c
dA
dA
1 + A
pA
Each price pi , for i = O, S, A, is determined by the standard markup rule (i.e., the term
11
i i
ci 1+p i ), augmented to take into account the impact of a change in pi on the other two groups
pi
of consumers. Observe that the ability for the newspaper to price discriminate between
subscribers and occasional readers leads to prices being chosen as if there were three distinct
sides to this industry—subscribers, occasional readers, and advertisers—where the number of
active participants in each each side of the industry impacts the other two sides. Specifically,
the number of advertisers impacts the readers’ payoffs, and similarly the number of readers
impacts the advertisers’ payoffs. However, in addition to these standard effects, the price
charged to occasional readers affects the number of subscribers, and the price charged to
subscribers affects the number of occasional readers. In other words, when a platform in a
two-sided industry can engage in second-degree price discrimination within one side of the
industry, the optimal pricing rules are qualitatively identical to those that would emerge if
the industry had three distinct sides.
Observe also that the newspaper would engage in second degree price discrimination even
if there were no advertisers (to see this, set dA = 0 in the formulas of pS and pO ). Readers
with a high taste for reading would buy every issue of the newspaper at the newsstand at
price pO (even when t = 0) if subscribing was not possible. Because subscribing is instead
possible, and because pS < pO , these readers prefer subscribing to enjoy the lower average
price. Readers with an average taste for reading instead have a low enough gross payoff when
t = 0 that it is not interesting for them to have all n issues be delivered to their home; they
prefer buying it only when t = x, even though the per issue price p is higher. Here lies
the scope for price discrimination. Setting pS < pO means extracting less surplus from the
readers with a rather high taste for reading (those who would have bought the newspaper at
the newsstand anyway), but allows the platform to extract more surplus from the informed
consumers; i.e., those who buy only when t = x. It is thus these informational differences
that the newspaper exploits through second-degree price discrimination. In other words, it
is not the presence of advertisers that explains the existence of price discrimination in this
model; though advertisers will certainly affect its extent.
2.3
Explaining the empirical results
One empirical result we have that our model explains well is the following. We observe that as
newspapers decrease the subscription price (and leave unchanged the per unit price), the share
of subscribers increases but the total number of readers is little affected. This is completely
in line with our model of price discrimination. In our model, the readers indifferent between
subscribing and being occasional readers base their decision on both pO and pS . In particular,
there exists a threshold θ̃1 pO , pS such that all readers whose taste for reading is larger than
θ̃1 subscribe, and the rest do not. Moreover, within the set of readers who do not subscribe,
their purchasing decision depends only on pO . In other words, there exists a threshold θ̃0 pO
12
h
i
such that all readers whose taste for reading belongs to θ̃0 pO , θ̃1 pO , pS
sometimes
O
buy the newspapers, and those whose taste for reading is below θ̃0 p never purchase the
newspaper. As is clear, a change in pS alone affects the composition of readers between
subscribers and non subscribers, but not the total number of readers.
More generally, obtaining the result that the newspaper readjusts towards more subscribers
following the shock is not conceptually difficult to understand (independently of whether this
is achieved by increasing or decreasing prices). It must be the case, prior to the shock,
that marginal advertisers – those indifferent between placing an ad or not – were putting
a higher weight on occasional buyers than the newspaper. Following the negative shock on
advertisers – which, inevitably, leads to the newspaper catering less to what advertisers want
– the newspaper readjusts its readership towards more subscribers. This effect is reinforced
in case the newly marginal advertisers – i.e., following the shock – also put more weight on
subscribers than occasional readers. But even if the latter does not hold, we expect the first
force to dominate.
Obtaining the result that prices charged to readers decrease following a negative shock on
the advertisers’ side is not straightforward. This is because there is a very strong force that
suggests that, following the shock, the newspaper should pay less attention to what advertisers
want (i.e., more readers), since it can make less money out of them. Several explanations can
however be advanced.
First, there is also a negative shock on the readers side (in addition to that on the advertisers’ side), which leads to the newspaper decreasing prices to readers. This, however,
does not seem like the right explanation in our case, because it seems hard to believe that the
shock was more pronounced on the readers’ side than on the advertisers’ side. There was no
change in the number nor in the quality of the programs broadcasted on Television after the
introduction of advertisement.
Second, if we assume that (i) readers dislike ads and (ii) advertisers dislike (other) ads,
an explanation is possible. Prior to the shock, the advertisers’ willingness to pay for a slot in
the newspaper is high, and higher the lower the number of other ads. As a result, prior to
the shock, the newspaper is in the best of all possible worlds: it is able to put few ads in its
pages (which is good for readers) and charge a premium for it. Readers like the fact that so
few ads are present, and are thus also charged a high price. Following the negative shock on
advertisers, it could be that advertisers are no longer willing to pay a lot for “exclusivity”,
so that the newspaper stops granting them exclusivity and increases the quantity of ads as a
result. Because readers dislike ads, they must be charged a lower price to be compensated.
Third, we could envision a competition effect. When television starts competing for advertisers with newspapers, newspapers have no choice but to offer a “better” product to
advertisers, that is, more readers (which is achieved by setting lower reader prices). Writ-
13
ing down a model that replicates this intuition is not immediate, as increased competition
depresses ad prices, and thus lowers the incentives for newspapers to cater to advertisers’
preferences.
3
Industry and data characteristics
In this section, we briefly introduce the new dataset we built for this study and describe the
newspaper industry characteristics. We discuss further details of the construction of the data
in the online Appendix Section A.
3.1
Newspaper industry characteristics
The French daily newspaper industry is divided into two sub-industries: the local daily newspaper industry and the national daily newspaper industry. During our period of interest
(1960-1974), there are around 100 (national and local) daily general information newspapers.
There are 12 national newspapers at the beginning of the period and 10 at the end.11
The total national newspaper circulation is stable during this time period, with around 4.2
million copies sold every day. The number of local newspapers during the same period varies
around 90, with a total circulation amounting to around 7.8 million copies (see Cagé (2014)
for more details on the historical evolution of the French local daily newspapers industry).
On average, the circulation of national daily newspapers amounts to nearly 350,000 copies a
day and the one of local daily newspapers to 100,000. Copies are sold either at the newsstand
to unit buyers or through subscription. The average share of unit buyers is 70%.12 As
expected, the price charged to subscribers is lower than the price paid by unit buyers at
the newsstand. The average price ratio is 0.86. (Table 1 provides descriptive statistics on
newspaper prices, revenues and costs as well as on circulation and newspaper content for the
entire daily newspaper industry.13 )
Overall, national daily newspapers generate 67.5 million francs (e71.4 million) in total
revenues each year, and local daily newspapers 19.9 million francs (e20.4 million). These
revenues come from sales and from advertising. On average, between 1960 and 1974, the
share of advertising revenues in total revenues is 45%. The quantity of advertisement in
newspapers represents around 3 pages per newspaper issue, i.e., 19% of the content of the
newspaper.
11
Libération and Paris Presse exit the industry respectively in 1964 and in 1970.
The statistical discrepancy between the share of unit buyers and the share of subscribers – that do not
sum up to 100 – stems from the fact that a number of copies is distributed for free every day.
13
In the online Appendix, we present these descriptive statistics separately for national – Table C.1 – and
local – Table C.2 – daily newspapers.
12
14
15
0.40
0.35
0.87
8.50
10.12
4.19
5.51
45
9.49
39
0.37
57,655
75
22
307
15
12
3
17
26.51
12.89
13.53
45
25.67
65
0.85
135,387
70
27
307
16
12
3
19
Median
0.43
0.38
0.86
11.95
Mean
7
4
4
10
178,522
23
22
12
38.04
67
6.21
38.40
22.03
18.45
11
0.20
0.17
0.07
11.06
sd
2
2
0
2
1,725
2
1
218
0.07
1
-36.89
0.10
0.06
0.03
6
0.10
0.08
0.17
0.75
Min
66
34
32
62
1,143,676
100
98
365
261.37
326
41.42
247.12
181.27
123.87
82
2.40
0.94
1.25
50.00
Max
996
996
997
996
973
971
968
970
958
840
952
962
968
964
962
1,043
967
967
662
Obs
Notes: The Table gives summary statistics. Time period is 1960-1974. Variables are values for newspapers. The observations are at the newspaper/year level.
Unit price and subscription price per issue are in francs. Revenues and costs are in million francs.
Prices
Unit buyer price
Subscription price
Price ratio
Display ad rate (listed price)
Revenues
Total revenues
Revenues from advertising
Revenues from sales
Share of advertising in total revenues (%)
Expenditures
Total expenditures
Number of journalists
Profit
Circulation
Total circulation
Share of unit buyers (%)
Share of subscribers (%)
Frequency (issues/52weeks)
Content
Number of pages
News hole (nonadvertising space)
Advertising space
Share of advertising
Table 1: Summary statistics: Newspapers
3.2
County demographics and economic data
While national daily newspapers circulate (by definition) in the entire country, the natural
news market for a local daily newspaper in France is a county.14 A number of newspapers
circulate across nearby counties, however.15
Table 2 presents summary statistics on county demographics and economic data (with
information by county on the total number of full time workers, the total number of firms and
the aggregate corporate tax).
3.3
Data
We collect an annual balanced panel dataset on local and national newspapers in France
between 1960 and 1974. The data is paper data that we digitize and merge from the French
Ministry of Information’s non-publicly available records in the National archives. Newspapers
were asked by the Ministry of Information to report annually on revenues and expenses. We
collect data by having direct access to their responses to these queries.
Local and national newspapers Our dataset includes data for 61 of the local newspapers,
i.e., more than three quarters of the local daily newspapers industry in 1971. These newspapers are the only ones for which the data is available in the archives. They represent on
average more than 87% of the total local daily newspaper circulation. Our sample of national
newspapers include all the 10 national newspapers circulating between 1960 and 1974.
Price, cost and revenue data
For the 71 newspapers described above we collect data on
prices with information on unit price, subscription price, and the number of issues per year.
This allows us to compute a measure of price discrimination. We also have data on revenues
(from sales and from advertising), and on costs, as well as data on the number of journalists.
Finally, we have data on circulation with the share of unit buyers and the share of subscribers.
Advertising prices and quantity A change in advertising revenues can be driven by
either or both a change in advertising prices or a change in advertising quantity. We collect
data on both the price and the quantity of advertising in order to disentangle the two effects.
A first source of information for advertising prices are the listed prices per advertising slot.
We digitize this data from “Tarif Media”, an annual publication that provides information
regarding advertising prices. However, a downside of using listed prices is that discounts are
common in this industry. Price lists are hence a relevant measure of advertising prices as long
14
A county (“dpartement” in French) is a French administrative division. The median land area of a county
is 2,303 sq mi, which is slightly more than three-and-half times the median land area of a county in the United
States. There are 95 counties in metropolitan France.
15
In 1968, 54% of the local daily newspapers circulate in only one county; 12% in two.
16
17
66.9
6.5
3.5
10,536
92
389,516
14,202
53,165
24
67.1
6.9
4.0
11,086
192
516,023
19,810
95,586
110
Median
25,233
172,996
613
3.8
3.4
1.9
3,060
470
420,726
sd
Notes: The Table gives summary statistics. Time period is 1960-1974. The observations are at the county/year level.
County demographics
Percent of population over 20
Percent of population with secondary education or higher
Percent of senior executives and knowledge workers
Average annual wage (in francs)
Income tax (in million francs)
Population
Economic data
Number of firms
Number of full time workers
Corporate tax (in million francs)
Mean
2,985
3,289
0
57.0
1.7
0.2
5,736
4
75,173
Min
Table 2: Summary statistics: County demographics and Economic data
262,854
2,690,000
9,438
98.4
25.6
16.4
25,156
5,200
2,856,531
Max
455
912
824
1,423
1,407
1,425
912
824
1,425
Obs
as we assume that the potential bias between list prices and actual prices does not differ too
much across newspapers and over time.
Given this caveat, we use another measure of advertising price common in the literature,
which consists of the total advertising revenues divided by the newspaper circulation. The two
measures are strongly correlated (the correlation between them is equal to .5 and is significant
at the 1% level).
We collect data on the amount of advertisement per issue directly from the paper version
of the newspapers available in the French National Library. For each year and each newspaper, we study the content of the newspaper issues during an entire week (the third week of
March16 ). We measure the quantity of advertisement on each page. We thus have information
on the total amount of advertisements in the newspaper, and on the share of the newspaper
that is devoted to advertisement.
Finally, for a subset of newspapers, we also collect information on the type of advertisements in the newspapers and obtain information on the category of each advertisement (e.g.
alimentation, automobile, household electrical goods,...) as well as on the brand advertising
in the newspaper.
4
Background on the introduction of advertisement on French
Television
The model we built in the previous section provided us with a general framework with which
to think about the determinants of pricing policies by newspapers, including the extent of
price discrimination. In this section, we study empirically how price discrimination varies
with advertising revenues. The empirical strategy we follow is in the spirit of an event study.
We exploit the introduction of advertisement on French Television in October 1968 as an
exogenous negative shock on the advertising side of newspapers. To the best of our knowledge,
we are the first to use this quasi-natural experiment.
In this section, we first present some historical background on the introduction of advertising on French Television in 1968, and then document a substitution effect from advertisement
in national newspapers to advertisement on television.
4.1
French Television in 1968
French Television is state-owned from 1945 to 1981.17 A national agency – the “Office de
Radiodiffusion-Télévision Française” (ORTF) – is in charge of providing radio and television
16
The choice of the third week of March was dictated by the fact that this is the week used by the INSEE
to run all its surveys.
17
During this period all TV channels are privately-owned in the US, while in the UK two TV channels are
state-owned (BBC 1 and BBC 2) and one is private (ITV).
18
content.18 Only one channel (“La première chaı̂ne” – the “First Channel”) is available until
1963. A second TV channel (“La deuxième chaı̂ne” – the “Second Channel”) is introduced in
1964 and a third one (“La troisième chaı̂ne” – the “Third Channel”) in 1972. TV penetration
is increasing at the time: in 1970, nearly 70% of the French households own a television
(Parasie, 2010).
Channels are financed mostly through a fee (redevance) until 1968. By law, commercial or
brand advertising is forbidden, with the exception of “collective advertising”. Collective ads
promote products, say fruits, without mention of a brand.19 They were not very important
however. In 1959 for example, the time devoted to collective advertising is only of five hours
and ten minutes per year (Parasie, 2010).
The transition to color on the Second Channel and the need to produce an increasing
number of programs means that the ORTF experiences severe financial difficulties – it is “on
the edge of the abyss” (Bellanger, 1969).20 Secretly decided by the French Government in
March 1965, the introduction of advertisement on television is made public on October 20th
1967, thereby provoking a strong controversy both in Parliament and within the Newspaper
Industry. The then Prime Minister George Pompidou argues that the ORTF has no choice
but to find new sources of revenues to develop the Second Channel and eventually create a
third one. He also argues that the introduction of advertisement on television will “revitalize
the production by giving to our firms the possibility to develop their domestic market, essential
support to any exporter activity.” (address in Parliament on April 24, 1968).21
4.2
A threat to newspapers?
Left-leaning political parties and the newspaper industry were firmly against the reform.
The Federation of the Democratic and Socialist Left (“Fédération de la gauche démocrate et
socialiste”) – a conglomerate of French left-wing non-Communist forces – introduced various
bills to ban commercial advertising on television by arguing that it would lead to a decrease in
the quality of television content. More importantly – and consistently with the identification
strategy we use in this paper –, very much present is the idea that the reform would lead
to a decrease in newspaper advertising revenue.22 In fact, already in 1964, the Minister
18
The first national agency, the “Radiodiffusion Française” (RDF), is created in 1945. It is eventually
renamed “Radiodiffusion-Télévision Française” (RTF) in 1949 and replaced by the ORTF in 1964.
19
These are allowed since the 1950’s and are also referred to as “compensatory advertising” (“publicité
compensée”), where the term“compensatory” captures the fact that the ORTF would receive a compensation
in exchange for the broadcast (Duchet, 2005). Not only advertisers have to constitute associations, but an
advertising campaign also needs the approval of the supervisory Ministry (e.g., the Ministry of Agriculture for
oranges).
20
Beginning on October 1st 1967, the Second Channel broadcasts twelve hours a week of programs in color.
21
Commercial advertising is allowed much earlier in almost all other developed countries: it is allowed in
1941 in the US, in 1955 in the UK, in 1956 in Germany, and in 1957 in Italy and Spain (Parasie, 2010).
22
The Federation of the Democratic and Socialist Left argues that the government wishes to introduce
advertising on television so as to weaken newspapers, the only independent media (Parasie, 2010). In an
19
of Information of the time, Alain Peyrefitte, was aware of this issue and claimed that the
introduction of advertising on television would be worth considering only if the press could
survive it (Bellanger, 1969).
Newspapers were similarly against the reform as they anticipate a decrease in their advertising revenues. And indeed, as underlined by Bellanger (1969), “in terms of national
advertising (...) in a limited market, any drain leads to a decrease in the advertising revenues
which the press lives off”. The Federation and the Confederation of the French Press estimated in a report that the press would lose between 40 and 50% of its advertising revenues,
i.e., between 20 and 40% of total revenues depending on the newspaper.
4.3
A substitution effect
Despite these strong resistances from the newspaper industry and the opposition, the first
commercial advertisement is broadcasted on French Television in October 1968. The time
devoted to advertising is of 2 minutes per day in 1968 – only on the First Channel –, 4
in 1969, 8 in 1970 (i.e. 2,720 minutes per year) – year in which advertising is introduced
on the Second Channel, and more than 12 in 1971 (Bellanger, 1969). Advertising revenues
generated by the ORTF increase by 69 million francs (77 (constant 2009) million euros23 )
between 1967 and 1968 and by 197 million francs (e201 million) between 1968 and 1969. In
1971, advertising revenues represent 22% of the ORTF total revenues (Bellanger, 1969). Did
this increase lead to a symmetric decrease in newspaper advertising revenues?
In order to provide a sense of the effect of the introduction of advertisement on television
on the advertising revenues of local and national daily newspapers, we first provide aggregate
evidence at the industry level.24 Total advertising revenues of national daily newspapers
decrease by 21 million francs (e45 million) between 1967 and 1968, and then stabilize around
500 million francs. While the advertising market is expanding in France between 1967 and
1974, national newspapers advertising revenues are actually decreasing. On the contrary, local
newspaper advertising revenues increase during the same period (Figure 2). Moreover, the
share of national daily newspapers in total advertising revenues decreases from 14% in 1967
to 11% in 1974, as shown in Figure 3.
The introduction of advertisement on television in 1968 can thus be considered as a significant negative shock on the advertisers’ side of the national newspaper industry. Why did it
address to the Parliament on April 24 1968, Jacques Chambaz (from the Communist Party) claims that “the
introduction of commercial advertisement on television is but a new way to deal a blow to the broadsheet
newspapers that you consider not docile and flexible enough.”
23
In the remainder of the paper, to save on space, we simply use the terminology “euros” when refering to
“constant 2009 euros”.
24
In Section 5, we provide econometric evidence of this shock, computing differences-in-differences estimates
to show that this shock affects negatively the advertising revenues of the national daily newspapers, but not
those of the local daily newspapers.
20
1,000
Million euros (constant 2009)
400
600
800
200
0
1967
1974
National news
1967
1974
1967
Local news
1974
Television
Notes: The Figure shows for 1967 and 1974 the value of advertising revenues in France by media outlets (local and
national daily newspapers, and total) in million euros (constant 2009). Data is from the “Institut de Recherches et
d’Etudes Publicitaires” (IREP), a French research institute devoted to the study of advertising.
Figure 2: Advertising revenues by media outlets, 1967 & 1974
1967
1974
11%
14%
20%
21%
12%
National daily newspapers
TV
Radio
Cinema
Local daily newspapers
Magazines
Outdoor
Others
Notes: The Figure shows for 1967 and 1974 the share of total advertising revenues representend by type each media
outlet (national daily newspapers; local daily newspapers; magazines; television; radio; cinema; outdoor; and others).
Data is from the “Institut de Recherches et d’Etudes Publicitaires” (IREP), a French research institute devoted to the
study of advertising.
Figure 3: Share of total advertising revenues by media outlets, 1967 & 1974
21
1,500
Number of advertisements
500
1,000
0
1968
1969
1970
1971
1972
1973
1974
Notes: The Figure shows the total number of new advertisements broadcasted every year on French Television between
1968 and 1974.
Figure 4: Number of new advertisements broadcasted on Television, 1968-1974
mainly affect national newspapers? Because the nature of advertising varies between national
and local newspapers. In particular, advertisements in national newspapers are mostly commercial advertisements that are relatively close substitutes to those broadcasted on television,
while a large share of advertisements in local newspapers are local in nature (local commercial
advertisements and classified advertisements).
4.3.1
Classifying advertisements
To document the substitution effect from advertisements published in national newspapers to
advertisements broadcasted on television, we created 25 advertisement categories: food and
drink, automobile, household electrical good, clothing,...
Television
We collect data on all the advertisements broadcasted on French Television
between 1968 and 1974 from the website of the Institut National de l’Audiovisuel (INA –
National Audiovisual Institute). For each advertisement, we have information on the exact
date of its first diffusion, its length as well as its category. Between 1968 and 1974, 7,337
different advertisements were broadcasted on television (142 in 1968, 919 in 1969 and more
than 1,000 per year for every following year, as shown in Figure 4). Online Appendix Figure
B.4 illustrates the prevelance of certain kind of categories compared to others on television
(e.g. 34% of all the advertisements broadcasted on French television in 1971 were for food
and non-alcoholic drink).
22
Newspapers In order to compare the advertisements broadcasted on television with the
ones published in newspapers, we similarly classify all the advertisements published in newspapers by categories. More precisely, for a subset of newspapers (four national newspapers25
and five local newspapers26 ), we classify all the advertisements published in the newspaper
between 1963 and 1972. To do so, we use the same method as the one described above for
the quantity of advertising (third week of March). Moreover, for each advertisement in newspapers, we classify it as local or not. According to our findings, 24% of the advertisements in
national newspapers are local advertisements, much less than in local newspapers where they
represent 44% of total advertisements.27
4.3.2
Anecdotal evidence
We document the existence of a substitution effect between advertisements in national newspapers and advertisements on television. Between 1960 and 1974, this substitution pattern
appears very clearly, as illustrated in Figure 5.
5
Empirical analysis
5.1
Estimation strategy
We use our panel data to compute differences-in-differences (DD) estimates of the effect of the
introduction of advertisement on television. Our identifying assumption is that the negative
shock on advertising revenues has affected mostly national daily newspapers, and to a lower
extent local daily newspapers. Indeed, as highlighted above, while national newspaper advertisement consists mostly of commercial advertisements that are relatively close substitutes to
those broadcasted on television, a large share of advertisements in local newspapers is instead
local in nature. Hence national newspapers are more likely to respond to the negative shock on
national advertising than local newspaper. We thus use national newspapers as our “treated
group”, and local newspapers as our “control group”, and compare the pre-1968-to-post-1968
change in prices of national daily newspapers to the change in prices of local daily newspapers
over the same period.
Let Dnational news be an indicator variable for national newspapers and Dafter be a time
dummy that switches on for observations post 1968 (i.e., after the introduction of advertisement on television). Our analysis is based on the following regression equation:
25
France Soir, L’Aurore, Le Figaro and Le Monde
La Liberté De Normandie, La Marseillaise, Le Maine Libre, Le Méridional and Le Midi Libre.
27
These estimates are consistent with existing aggregate data on revenues: according to IREP, the share of
local advertisements in advertising revenues of local daily newspapers was equal to 43% in 1967.
26
23
1965
1966
1967
1968
year
1969
1970
1971
40
30
20
10
1972
0
Total number of advertisements on TV
5
4
3
2
1
10
0
1964
Electronic devices and computer hardware (share)
0
Share of advertisements in National Newspapers (%)
30
20
3
2
1
0
Number of advertisements on TV
4
40
Number of advertisements in National Newspapers
Electronic devices and computer hardware (number)
1964
1965
1966
1967
1968
year
1969
National Newspapers (number)
National Newspapers (share)
Television
Television
1970
1971
1972
OTC Drugs (share)
1965
1966
1967
1968
year
1969
1970
1971
1972
10
5
0
1964
1965
1966
1967
1968
year
1969
National Newspapers (number)
National Newspapers (share)
Television
Television
(c) OTC drugs (number)
Total number of advertisements on TV
3
2.5
2
1.5
1
.5
0
15
10
5
0
Number of advertisements on TV
4
3
2
1
0
1964
15
OTC Drugs (number)
Share of advertisements in National Newspapers (%)
(b) Electronic devices and computer hardware
(share)
Number of advertisements in National Newspapers
(a) Electronic devices and computer hardware
(number)
1970
1971
1972
(d) OTC drugs (share)
Figure 5: Anecdotal evidence: substitution from advertisements in national newspapers to
advertisements on television
24
yn,t = α + β1 Dafter + β2 (Dafter ∗ Dnational news )
(12)
+λn + γt + n,t
where n indexes newspapers and t indexes years (t = 1960, ...1974). For all specifications in
our analysis, we introducte fixed effects for each newspaper (λn ) as well as time dummies (γt ).
This prevents cross-sectional variations from driving our results. n,t is a newspaper-countyyear shock.
yn,t is our outcome of interest. We first consider the log of the price ratio (defined as the
subscription price per issue divided by the unit price) as our dependent variable. We assume
that the difference in prices charged to unit buyers and subscribers is entirely due to price
discrimination and use the price ratio as our measure of price discrimination (Clerides, 2004).
Obviously, part of the difference between the prices charged to unit buyers and subscribers
may be driven by differences in costs, in particular costs of delivery. However, our assumption
is valid in the DD setting as long as the introduction of advertisement on television did not
affect costs of delivery.
Due to the inclusion of newspapers and year fixed effects, the coefficient β2 – our coefficient
of interest – measures the annual price ratio effect for national newspapers of the introduction
of advertisement on television compared to the general evolution of the price ratio for local
newspapers. The key identifying assumption here is that price trends would be the same for
both categories of newspapers (local and national) in the absence of treatment. The treatment
induces a deviation from this common trend. Figure 6 provides strong visual evidence of
treatment and control newspapers with a common underlying trend, and a treatment effect
that induces a sharp deviation from this trend. Note that, given local newspapers may also
have suffered from the shock – but to a lower extent given their lower reliance on national-ad
revenue – our estimates are a lower bound.
Finally, the unbiasedness of the DD estimates requires the strict exogeneity of the introduction of advertisement on television. As we underline above, French Television is state-owned
from 1945 to 1981. There is thus no interaction between television owners and newspaper
owners, be they national or local. The introduction of advertising on television was decided
unilaterally by the French government to answer the concerns of the ORTF. It is exogenous
to the newspaper industry.
5.2
5.2.1
Results
The effect on prices
Table 3 reports estimates of equation (14). Our outcomes of interest are the different prices
charged by newspapers: the price ratio (column 1), the subscription price (column 2), the unit
25
.9
.85
Price ratio
.8
.75
1974
1973
1972
1971
1970
1969
1968
1967
1966
1965
1964
1963
1962
1961
1960
.7
Year
Average price ratio - National newspapers
Average price ratio - Local newspapers
Figure 6: Descriptive evidence: Changes in price discrimination
buyer price (column 3) and the advertising price (columns 4 and 5). We find a 10% decrease
in the price ratio – our measure of price discrimination – of national newspapers compared
to local newspapers following the introduction of advertisement on television. This decrease
is statistically significant at the 1% level.
This decrease is driven by a decrease in the subscription price (column 2): following the
introduction of advertisement on television, the subscription price decreases by 12%. There
is no statistically significant change in the unit price.
Regarding advertising prices, we find a 28% decrease following the shock, whether we use
our listed price measure of advertising prices (column 5), or the total advertising revenues
normalized by circulation (column 4). This decrease is consistent with the view that the
introduction of advertisement on television can be considered as a negative shock on the
advertising side for national daily newspapers.
The equation errors may be correlated in this set of price equations. We thus estimate the
system of prices using seemingly unrelated regression (SUR) analysis. We find that the errors
are correlated. Table 4 presents the results. Similarly to the results in Table 3, we find that
the introduction of advertisment on television leads to a decrease in the subscrition price and
in the advertising price of national newspapers compared to local newspapers. The results
are of similar order of magnitude than the one we obtain above, and statistically significant
at the 1% level.
26
Table 3: Baseline estimation: Prices
Advertising prices
Post-1968
National x Post-1968
R-sq
Observations
Price ratio
-0.01
(0.02)
-0.12∗∗∗
(0.01)
0.46
967
Subscription price
1.29∗∗∗
(0.01)
-0.12∗∗∗
(0.01)
0.98
967
Unit price
1.30∗∗∗
(0.02)
-0.01
(0.02)
0.98
967
Ad revenues / circulation
1.36∗∗∗
(0.04)
-0.25∗∗∗
(0.06)
0.91
968
Listed price
0.48∗∗∗
(0.13)
-0.25∗∗∗
(0.06)
0.89
662
Notes: * p<0.10, ** p<0.05, *** p<0.01. Standard errors in parentheses are robust. Time period is 1960-74.
Models are estimated using OLS estimations. All the estimations include newspaper and year fixed effects.
Variables are described in more details in the text.
Table 4: Seemingly unrelated regression: Prices
Subscription price
Post-1968
National x Post-1968
Unit price
Post-1968
National x Post-1968
Ad revenues / circulation
Post-1968
National x Post-1968
Subscription price
Subscription price
0.77∗∗∗
(0.01)
-0.12∗∗∗
(0.01)
1.26∗∗∗
(0.07)
-0.13∗∗∗
(0.02)
1.05∗∗∗
(0.01)
-0.01
(0.01)
1.03∗∗∗
(0.08)
-0.02
(0.02)
1.36∗∗∗
(0.04)
-0.25∗∗∗
(0.04)
Listed price
Post-1968
National x Post-1968
R-sq
Observations
0.98
964
1.70∗∗∗
(0.38)
-0.26∗∗∗
(0.09)
0.97
635
Notes: * p<0.10, ** p<0.05, *** p<0.01. Time period is 1960-74. Models are estimated using Zellner
Seemingly Unrelated Regression technique. All the estimations include newspaper and year fixed effects.
Variables are described in more details in the text.
27
Table 5: Baseline estimation: Circulation
Number unit buyers
Post-1968
National x Post-1968
∗∗
-0.09
(0.04)
-0.01
(0.05)
Unit price
Subscription price
R-sq
Observations
0.99
971
-0.00
(0.09)
-0.01
(0.04)
-0.11
(0.09)
0.02
(0.10)
0.99
966
Number subscribers
Share subscribers
-0.10
(0.06)
0.14∗∗
(0.06)
-0.09
(0.06)
0.16∗∗∗
(0.06)
0.97
968
-0.11
(0.16)
0.13∗∗
(0.06)
0.10
(0.09)
-0.10
(0.11)
0.97
963
0.96
968
-0.20
(0.14)
0.15∗∗
(0.06)
0.17∗∗
(0.09)
-0.10
(0.10)
0.96
963
Notes: * p<0.10, ** p<0.05, *** p<0.01. Standard errors in parentheses are clustered by newspaper. Time
period is 1960-74. Models are estimated using OLS estimations. All the estimations include newspaper and
year fixed effects. Variables are described in more details in the text.
5.2.2
The effect on circulation
To which extent does this change in price discrimination affect the demand for newspapers?
We estimate the impact of the introduction of advertisement on television on the number
of unit buyers and the number of subscribers. Table 5 presents the results. We obtain a
statistically significant increase in the number of subscribers which goes up by 13 to 14%
(columns 5 and 6). The number of unit buyers stays unchanged; as a result the share of
subscribers in total circulation increases as a consequence of the shock.
5.2.3
The effect on revenues
Hence the introduction of advertisement on television negatively affected the subscription
price and the advertising price of national newspapers compared to local newspapers, but it
leads to an increase in the number of subscribers. We then turn to its effect on revenues
(both from sales and from advertising), as well as on profit. Table 6 presents the results. We
find that the shock leads to a 13% decrease in total revenues. This decrease comes both from
a decrease in the revenues from sales (of 10%) and from a more important decrease in the
revenues from advertising (of 27%).28 As a result, the advertising share in total revenues also
decreases after the shock. This is also the case of newspaper profits which fall by 45%.
The decrease in advertising revenues comes from the decline in the price of advertising. In
Table 7, we present estimations of the effect of the introduction of advertisement on television
on the content of newspapers. We find no statistically significant result; in particular, the
number of advertising pages in newspapers stays constant. Hence as a consequence, with a
28
As for prices, the equations may be related through the correlation in the errors. We thus estimate the
effect on sale and advertising revenues using SUR. Table D.1 in the online Appendix shows that our results
are robust to this alternative estimation strategy.
28
Table 6: Baseline estimation: Revenues
Revenues
Post-1968
National x Post-1968
R-sq
Observations
Total
1.12∗∗∗
(0.04)
-0.13∗∗∗
(0.05)
0.99
962
From sales
1.22∗∗∗
(0.04)
-0.10∗∗∗
(0.04)
0.99
964
From advertising
1.34∗∗∗
(0.06)
-0.27∗∗∗
(0.07)
0.98
968
Advertising share
0.14∗∗∗
(0.03)
-0.14∗∗∗
(0.04)
0.86
962
Profit
0.91∗∗∗
(0.33)
-0.45∗
(0.24)
0.74
627
Notes: * p<0.10, ** p<0.05, *** p<0.01. Standard errors in parentheses are robust. Time period is 1960-74.
Models are estimated using OLS estimations. All the estimations include newspaper and year fixed effects.
Variables are described in more details in the text.
Table 7: Baseline estimation: Content
Post-1968
National x Post-1968
R-sq
Observations
Number of pages
0.42∗∗∗
(0.03)
0.00
(0.03)
0.94
996
News hole
0.35∗∗∗
(0.03)
-0.03
(0.03)
0.91
996
Advertising space
0.62∗∗∗
(0.07)
-0.01
(0.05)
0.87
996
Share of advertising
0.21∗∗∗
(0.06)
-0.00
(0.04)
0.74
996
Notes: * p<0.10, ** p<0.05, *** p<0.01. Standard errors in parentheses are robust. Time period is 1960-74.
Models are estimated using OLS estimations. All the estimations include newspaper and year fixed effects.
Variables are described in more details in the text.
smaller price and an unchanged quantity, advertising revenues decrease after the introduction.
5.3
Interpretation of the results
Our proposed interpretation of these empirical findings, consistent with the theoretical predictions of our model, is that while newspapers have a preference for subscribers, advertisers
prefer unit buyers. Figure 7 illustrates this preference: it plots the correlation between the
advertising price paid by advertisers and the share of unit buyers of the newspaper. We obtain a positive correlation: the higher the share of unit buyers (compared to subscribers), the
higher the price advertisers are willing to pay.
As a consequence, when advertising revenues play an important role, newspapers tend
to distort their pricing strategies in favor of unit buyers. On the contrary, when advertising
revenues fall (e.g. after the 1968 French reform), they revert to their own profit-maximizing
pricing strategies (independent of the advertising side of the market), i.e. they increase price
29
50
40
Advertising listed price
20
30
10
0
0
10
20
30
40
50
60
Share of unit buyers (%)
70
80
90
100
Figure 7: Descriptive evidence: Advertisers’ preference for unit buyers
discrimination in favor of subscribers.
5.4
Robustness
We perform several robustness checks. This section briefly describes them; the detailed results
for these tests are available in the online Appendix.
Industry-specific time trend As an alternative check on the DD identification strategy,
we add an industry-specific time trend to the list of controls. The introduction of this industryspecific time trends allows treatment and control newspapers to follow different trends in a
limited but potentially revealing way. Table E.1 in the online Appendix gives the results.
We find that the increase in price discrimination, driven by the decrease in the subscription
price, is robust to controlling for a national newspapers industry-specific time trend which is
reassuring as to the validity of our DD identification strategy.
Timing of the effect
The before-after event study approach enables us to control for time-
invariant newspaper-specific effects and general time trends. As an additional robustness
check, we allow for flexible time-varying effects of the negative shock on advertising revenues
(Laporte and Windmeijer, 2005). To quantify the dynamics effects of the event and control
for lags and leads, we define (“pulse”) variables for two, non-overlapping, three-years spaced
periods around the event and a dummy variable isolating the long-run effect of the shock (see
e.g. (Papaioannou and Siourounis, 2008)).
Our specification is:
30
log price ration,t = α + δ1 d1n,t + δ2 d2n,t + δ3 d3n,t
0
+Xn,t
∆ + λn + γt + n,t
(13)
where d1n,t = 1 in 1965, 1966, and 1967 for national newspapers (pre introduction of advertising
on television); d2n,t = 1 in 1968, 1969 and 1970 for national newspapers (at the time of the
introduction and in the following years); and d3n,t = 1 in 1971 and all subsequent postintroduction years (until 1974). Each indicator variable equals zero in all other years than
those specified and for local newspapers. Thus the base period is the years before 1965.
Online Appendix Table E.2 presents the results. We find no statistically significant effect
(with a point estimate close to zero) for the pulse variable d1n,t = 1, whether we consider the
price ratio, the subscription or the unit price. This is reassuring as to the validity of our
DD strategy. Moreover, as expected given the results of Table 3, we obtain a negative and
statistically significant at the 1% level δ2 : there is a statistically significant decrease in the
price ratio – i.e., in the extent of price discrimination – of national newspapers compared to
local newspapers following the introduction of advertising on television. This effect is long
lasting: the δ3 coefficient is statistically significant.
Sample selection For a small number of newspapers, we were not able to collect information for the entire period 1960-1974. Moreover, some newspapers in our sample exited the
market before 197429 . To check that our results are not affected by sample selection issues, we
compute the Difference-in-Difference estimation including in our sample only the newspapers
for which data is available for the entire period 1960-1974. Online Appendix Table E.3 gives
the results when the sample is restricted. The effects we obtain are statistically significant
and of the same order of magnitude.
Inflation
Finally, given the very high inflation rate in France in the 60’s and 70’s (up to
13% in 1974), we check that our results are robust to controlling for annual inflation. On
the one hand, we investigate the effect on constant rather than on current prices. On the
other hand, rather than introducing year fixed effects, we control for the annual inflation rate.
Online Appendix Table presents the results. They are not significantly different from those
presented in Table 3.
5.5
Reliance on advertising revenues
Finally, we consider an alternative within-local newspapers empirical strategy to investigate
the effect of the advertisers’ side of the industry on price discrimination. Newspapers differ
29
Three local newspapers – L’Echo de la Corrèze in 1971; L’Espoir de Nice which became a weekly newspaper
in 1973; and La Nouvelle Gazette de Biarritz which merged with another newspaper in 1972) – and two national
newspapers (Paris Jour in 1972 and Paris Presse in 1970 – exited the market before 1974.
31
widely on the extent to which they rely on advertising (as shown in the online Appendix
Figure B.3 which plots the distribution of the share of advertising in total revenues in 1967).
Hence they may have been affected differently by the shock.
Given that there are only 10 national newspapers, we cannot perform such a within analysis
on this industry. We are thus focusing on the local daily newspaper industry. We saw before
that local newspapers have been much less affected by the introduction of advertisement on
television than national newspapers; a large share of advertisements in local newspapers are
local in nature. However, this does imply that national advertisements are absent local daily
newspapers.
To investigate the effect of a drop in advertising revenues depending on the reliance on
advertising, we estimate the following regression equation:
yn,t = α + ν1 Dafter + ν2 (Dafter ∗ Dhigh reliance ) + λn + γt + n,t
(14)
where Dhigh reliance is an indicator variable equal to one for local daily newspapers whose
reliance on advertising revenues in 1967 was above the median, and 0 for those whose reliance
was below (for the consistency of the estimate, we drop the newspapers that are just around
the median).
Table 8 presents the results. The estimates we obtain are consistent with the ones using
the national vs. local newspapers empirical strategy. The magnitude of the coefficients is
lower (e.g. the price ratio only decreases by 4%) but they are statistically significant at the
5% level or better. This lower magnitude was expected given that we are performing here a
within-local newspaper estimation, and local newspapers have been much less affected by the
shock than national newspapers.
6
Conclusion
We have built a model in which a profit-maximizing newspaper must attract both readers
and advertisers. Particular attention has been paid to the incentives the newspaper has to
engage in second-degree price discrimination between subscribers and occasional buyers, and
how these incentives interact with the advertisers’ side of the industry, and in particular the
reliance on advertising revenues.
In our model, there is scope for second degree price-discrimination because of the readers’
uncertainty regarding their utility from reading in future periods. The newspaper sets its
prices such that readers with a high average taste for reading subscribe at a relatively low
unit cost, while readers with an intermediate average taste for reading only buy occasionally
– but at a high price – when their utility from reading on that day is high. One general
tendency that emerges is that, as long as advertisers prefer unit buyers to subscribers, price
32
Table 8: Reliance on advertising revenues: Within-local newspapers analysis
(a) Prices
Post-1968
High reliance x Post-1968
R-sq
Observations
Price ratio
-0.01
(0.04)
-0.04∗∗∗
(0.01)
0.36
549
Subscription price
1.30∗∗∗
(0.02)
-0.03∗∗
(0.01)
0.98
549
Unit price
1.32∗∗∗
(0.03)
0.01
(0.01)
0.97
549
Advertising price
1.44∗∗∗
(0.05)
-0.07∗∗
(0.03)
0.89
550
(b) Circulation
Post-1968
High reliance x Post-1968
R-sq
Observations
Number unit buyers
-0.09∗∗
(0.04)
-0.04
(0.03)
0.99
551
Number subscribers
-0.28∗∗∗
(0.08)
0.17∗∗∗
(0.04)
0.97
548
Share subscribers
-0.20∗∗∗
(0.08)
0.16∗∗∗
(0.04)
0.96
548
(c) Revenues
Post-1968
High reliance x Post-1968
R-sq
Observations
Total revenues
1.11∗∗∗
(0.04)
-0.05∗∗
(0.02)
0.99
546
From sales
0.97∗∗∗
(0.04)
-0.02
(0.02)
0.99
546
From advertising
1.37∗∗∗
(0.06)
-0.07∗∗
(0.03)
0.98
550
Profit
1.72∗∗∗
(0.39)
-1.34∗∗∗
(0.25)
0.71
368
(d) Content
Post-1968
High reliance x Post-1968
R-sq
Observations
Number of pages
0.44∗∗∗
(0.04)
0.01
(0.02)
0.96
551
News hole
0.39∗∗∗
(0.04)
0.01
(0.02)
0.93
551
Advertising space
0.64∗∗∗
(0.11)
-0.02
(0.06)
0.86
551
Share of advertising
0.19∗
(0.10)
-0.03
(0.06)
0.71
551
Notes: * p<0.10, ** p<0.05, *** p<0.01. Standard errors in parentheses are robust. Time period is 1960-74. Models are
estimated using OLS estimations. All the estimations include newspaper and year fixed effects. Variables are described
in more details in the text.
33
discrimination increases following a general increase in the outside option of newspapers. This
is consistent with the empirical evidence we obtain using French daily newspapers between
1960 and 1974.
This empirical finding has implications for the 21st century newspaper industry. In particular it sheds light on the observed current tendency for newspapers to favor subscriber-based
readerships through low subscription prices (and high newsstand prices). What our paper
suggests is that when newspapers are less capable of generating revenues from advertising, all
else equal, they tend to adjust their pricing policies on the readers’ side so as to increase the
share of subscribers. As advertising revenues continue declining, we should thus expect this
tendency to reinforce itself in the coming years.
Moreover, our results also shed light on a new phenomenon characterizing the newspaper
industry: the tendency of a number of newspapers in the United States to charge more for a
digital subscription than for a print subscription, despite that it is much more costly to deliver
a print than a digital subscription. The Greensboro News & Record is thereby charging $187.12
for a 7-day, 52-week print subscription but $215.40 for a digital one. Similarly, the Orange
County Register is offering digital access for $3.99 a week, or digital plus Sunday print for
$2.99 a week, and the New York Times’ Sunday print gives all-digital access at a price thats
usually cheaper than all-digital access itself.30 What is driving this new pricing strategy?
According to our results, this could be explained by the fact that a digital reader may be
worth less than a print reader (see e.g. Noam, 2009)31 . To the extent that newspapers may be
less capable of generating advertising revenues from a digital-only subscription than from a
print one, they may tend to adjust their pricing policies to increase the share of print readers.
Whether it is indeed the case will be the topic of future research.
30
31
http://www.niemanlab.org/2014/07/when-a-digital-subscription-costs-more-than-a-print-one/
“A print reader generates more than 20 times times as much in revenues than an online reader.” (p.440)
34
References
Anderson, S. P. and Coate, S. (2000). Market Provision of Public Goods: The Case of
Broadcasting. NBER Working Papers 7513, National Bureau of Economic Research, Inc.
Argentesi, E. and Filistrucchi, L. (2007). Estimating market power in a two-sided market:
The case of newspapers. Journal of Applied Econometrics, 22(7):1247–1266.
Athey, S., Calvano, E., and Gans, J. (2013). The Impact of the Internet on Advertising
Markets for News Media. Working Paper 19419, National Bureau of Economic Research.
Bellanger, C. (1969). Histoire générale de la presse française: De 1958 à nos jours. Histoire
générale de la presse française. Presses Universitaires de France.
Borenstein, S. (1991). Selling Costs and Switching Costs: Explain Retail Gasoline Margins.
RAND Journal of Economics, 20(3):344–365.
Borenstein, S. and Rose, R. (1994). Competition and Price Dispersion in the U.S. Airline
Industry. Journal of Political Economy, 103(2):653–683.
Busse, M. and Rysman, M. (2005). Competition and Price Discrimination in Yellow Pages
Advertising. Rand Journal of Economics, 36(2):378–390.
Cagé, J. (2014). Media Competition, Information Provision and Political Participation. Working paper.
Caillaud, B. and Nijs, R. D. (2014). Strategic Loyalty Reward in Dynamic Price Discrimination. Marketing Science, 33(5):725–742.
Carroni, E. (2015). Competitive Behaviour-Based Price: Discrimination among Asymmetric
Firms. Working Papers 1501, University of Namur, Department of Economics.
Clerides, S. (2004). Price Discrimination with Differentiated Products: Definition and Identification. Economic Inquiry, 42(3):402–412.
Dai, M., Liu, Q., and Serfes, K. (2012). Is the Effect of Competition on Price Dispersion
Non-Monotonic? Evidence from the U.S. Airline Industry. Working paper.
DellaVigna, S. and Malmendier, U. (2004). Contract Design and Self-Control: Theory and
Evidence. The Quarterly Journal of Economics, 119(2):353–402.
Duchet, C. (2005). Les années 70 : publicité & télévision. In La télévision en jeu, chapter 9.
CNRS.
35
Filistrucchi, L., Mangani, A., and Luini, L. (2012). Banning Ads from Prime-Time State TV:
Lessons from France. Working Papers 12-23, NET Institute.
Gabszewicz, J. and Sonnac, N. (1997). Subscription as a price discrimination device. CORE
Discussion Papers 1997090, Université catholique de Louvain, Center for Operations Research and Econometrics (CORE).
Gentzkow, M. (2014). Trading Dollars for Dollars: The Price of Attention Online and Offline.
American Economic Review, 104(5):481–488.
Gentzkow, M., Shapiro, J. M., and Sinkinson, M. (2011). The Effect of Newspaper Entry and
Exit on Electoral Politics. American Economic Review, 101(7):2980–3018.
Gentzkow, M., Shapiro, J. M., and Sinkinson, M. (2014). Competition and Ideological Diversity: Historical Evidence from US Newspapers. American Economic Review, 104(10):3073–
3114.
Gerardi, K. S. and Shapiro, A. H. (2009). Does Competition Reduce Price Dispersion? New
Evidence from the Airline Industry. Journal of Political Economy, 117(1):1–37.
Gil, R. and Riera-Crichton, D. (2011). Price discrimination and competition in two-sided
markets: Evidence from the Spanish local TV industry. IESE Research Papers D/894,
IESE Business School.
Glazer, A. and Hassin, R. (1982). On the economics of subscriptions. European Economic
Review, 19(2):343–356.
Godes, D., Ofek, E., and Sarvary, M. (2009). Content vs. Advertising: The Impact of Competition on Media Firm Strategy. Marketing Science, 28(1):20–35.
Grubb, M. D. (2009). Selling to Overcondent Consumers. American Economic Review,
99(5):1770–1807.
Grubb, M. D. (2012). Dynamic Nonlinear Pricing: biased expectations, inattention, and bill
shock. International Journal of Industrial Organization, 30(3):287–290.
Hagiu, A. (2009). Two-Sided Platforms: Product Variety and Pricing Structures. Journal of
Economics & Management Strategy, 18(4):1011–1043.
Jin, G. and Rysman, M. (2013). Platform Pricing at Sports Card Conventions. Working
Paper.
Kaiser, U. and Wright, J. (2006). Price structure in two-sided markets: Evidence from the
magazine industry. International Journal of Industrial Organization, 24(1):1–28.
36
Laporte, A. and Windmeijer, F. (2005). Estimation of panel data models with binary indicators when treatment effects are not constant over time. Economics Letters, 88(3):389–396.
Liu, Q. and Serfes, K. (2013). Price Discrimination in Two-Sided Markets. Journal of Economics & Management Strategy, 22(4):768–786.
Marcus Asplund Rickard Eriksson, N. S. (2008). Price Discrimination in Oligopoly: Evidence
from Regional Newspapers. The Journal of Industrial Economics, 56(2):333–346.
Morton, F. S. and Oster, S. (2003). Behavioral Decision-Making: An Application to the Setting of Magazine Subscription Prices. Working Paper 10120, National Bureau of Economic
Research.
Noam, E. (2009). Media Ownership and Concentration in America. Oxford University Press,
USA.
Papaioannou, E. and Siourounis, G. (2008). Democratisation and Growth. Economic Journal,
118(532):1520–1551.
Parasie, S. (2010). Et maintenant, une page de pub : une histoire morale de la publicité à la
télévision française, 1968-2008. M{é}dias essais. INA.
Resende, M. and Ferioli, E. (2014). Magazine Subscription and Intertemporal Discounting:
Some Further Evidence. Technical report.
Rochet, J.-C. and Tirole, J. (2003). Platform Competition in Two-sided Markets. Journal of
the European Economic Association, 1(4):pp. 990–1029.
Rysman, M. (2004). Competition between Networks: A Study of the Market for Yellow Pages.
The Review of Economic Studies, 71(2):pp. 483–512.
Seamans, R. and Zhu, F. (2014). Responses to Entry in Multi-Sided Markets: The Impact of
Craigslist on Local Newspapers. Management Science, 60(2):476–493.
Seim, S. and Viard, V. B. (2011). The Effect of Market Structure on Cellular Technology
Adoption and Pricing. American Economic Journal: Microeconomics, 3(2):221–251.
Weyl, E. G. (2010). A Price Theory of Multi-sided Platforms. American Economic Review,
100(4):1642–1672.
White, A. and Weyl, E. G. (2010). Imperfect Platform Competition: A General Framework.
Working Papers 10-17, NET Institute.
37